Production Management Experiences and Research at CIFE

Martin Fischer Director, CIFE and

Professor, Stanford University --

Nelly Garcia PhD Student, Stanford University

Introduction – CIFE

100% funded by A/E/C industries Building owners and developers Design and construction companies Software and hardware vendors

Timeline

1988 - 2000 | Building Information Modeling (BIM) 2001 - 2010 | Virtual Design and Construction (VDC) 2011 - pres | Facility Performance Optimization

The CIFE community invents the future construction practice

Practice

Education Research

CONSTRUCTION

CREATES

SOCIETY’S FIXED PHYSICAL WEALTH

We must optimize the performance of this

physical wealth

BUT Why do we come up short?

•  Unclear targets •  Uncoordinated workflows and

information

Other industries making “things” have increased the value added per work hour by 250% over the construction industry since 1964

Labor Productivity for construction industry vs. all non-agricultural industries

By Paul Teicholz, et. al., “US construction labor productivity trends, 1970 – 2008”

Integrated Concurrent Engineering (ICE)

Product Modeling (BIM+) Process Modeling

Current State Process, T5 Rebar Detailing for Construction

Des

ign

Man

ufac

ture

Engi

neer

ing

Release CADdwg, rebar

schedule (*.CSF)in Documentum

Documentcontrol delay

(1 week)

Preliminarydesign GA drawings

Refine RC detailsand concept for

buildability/detailing

Use model todevelop and

communicatemethods

Prepare RC detaildrawings(drafting)

Check andcoordinate detail

drawings

CAD check(1d/dwg)

Check againstengineering calcs

(.5d /dwg)

Independent finalcheck & sign off

(2 weeks)

Building controlcheck & sign off

(BAA, time?)

Release paper P4dwgs & bar

bendingschedules

Rebar factorystarts bending

Pre-assembly

Ship to site

Draft spec

Comment onspec

Update spec Release spec

Model rebarcomponent (Use

digitalPrototyping tool)

Issue and resolveTQ’s (Technical

Queries)

AutoCAD CAD RC IDEAS Arma +

Design input/changes

Technology:

Detailedengineering

designinformation

Site assembly

Preliminary drafting2 weeks

Back drafting1 week

Checking2 weeks

Document control1 weekTimeline:

NOTE: Drawings are batched into sections-then subdivided into building components.Each component is an assembly package,e.g. rail box floor, wall, etc.

The number of drawing sheets per buildingcomponent vary depending on the work. OnART for example, each component mayconsist of 8-15 GA drawings and 8-15 RCdetail drawings.

All of the GA drawings are complete -pend ing changes f rom other des igndisciplines

NOTE: Design changesduring detailing (from:architecture, baggage,s y s t e m s , e t c . ) a r eupsetting RC drawingdevelopment.

Other / None/Unknown

Preliminary RCdetailing

Iterativeprocess

Consists of:engineeringcalculations,sketches, etc.

Most of the checkingprocess is doneconcurrently with RCdetail development.

BAA building controlaccepts the opinionof the independentdesign check - anddoes not perform acheck of its own

Asse

mbl

y

Existing Process - 6 weeks

Client/Business Objectives

Project Objectives

Virtual Design and Construction (VDC)

High  Performance  building  

Integrated  Building  Systems  (Product)  

Integrated  Process  

Integrated  Team  

(Organiza>on)  

Integrated  Informa>on  

(BIM+)  

Value  (Metrics)  

Produc>on  Management  

Collabora>on  Coloca>on  

(ICE)  

Simula>on  Visualiza>on  

Team  Charter  /  Integrated  Form  of  Agreement  (IFOA)  

To achieve high-performance facilities, we need a strategy and methods for integration

Pay  for  integra>on  now  or  pay  for  it  later.  

Developed with Khanzode, Reed, and Ashcraft.

There are 3 types of work interdependencies: Pooled (independent) Sequential (dependent) Reciprocal (interdependent) From: Thompson, Organizations in Action, 1967

CIFE-SPS VDC Course at NCC, Helsinki Aug. 20-23, 2013

Feasibility

Concept

Design

Implementa-tion

Close Out

ICE

BIM (3D, 4D)

Process / Production Planning & Control

Metrics

Feasibility

Concept

Design

Construction

Startup

Feasibility

Concept

Design

Construction

Startup

Feasibility

Concept

Design

Construction

Startup

Feasibility

Concept

Design

Construction

Startup

Feasibility

Concept

Design

Construction

Startup

Include downstream knowledge in upstream decisions

Carry upstream decisions and information downstream

Con

side

r do

wns

trea

m k

now

ledg

e in

ups

trea

m p

hase

s

Consider upstream

information in dow

nstream phases

13

VDC

BIM

Prod Mgt

Metric

s

VDC Plan

BIM

ICE

Prod Mgt

VDC

VDC Plan

1-week intro

6-month implementation

2-day integration

Learning from Implementation

What’s next?

Clark Pacific

Don  Clark  (President,  Owner):  “By  digitally  modeling  all  the  parts  a  worker  touches  during  fabrica>on  and  erec>on  and  rethinking  the  produc>on  process,  we  were  able  to    •  increase  rebar  produc>vity  by  40%,  •  cut  tolerance  in  half,  •  reduce  rebar  waste  to  2%,  and  •  decrease  inventory  to  3  days.”  

First  ICE  Session  at  Graña  y  Montero  in  Lima,  Peru  

0  

57  

127  

286   287  287  

15  39  

97   97  

230  

0  

50  

100  

150  

200  

250  

300  

350  

0   10   20   30   40   50   60  

 Accum

ulated

 Que

ries  

Days   Issued  by  Graña  y  Montero   Answered  by  Client  

Issued  Queries  vs.  Answered  Queries  

Without  ICE  Session:  97  queries  answered  in  36  days  =  0.11  queries/hr  

ICE  Session:  66  queries  answered  in  8.5  hours  =  7.76  queries/hr  (1  query  every  8  min)  

189  WITHOUT  AN  ANSWER  

57  WITHOUT  AN  ANSWER  

ICE  SESSION  

ICE  Session  resulted  in  80x  increase  in  answers  to  queries    

Project Control Room by TAV-CCC-Arabtec JV on Midfield Terminal Project at Abu Dhabi Intl Airport

6  BIM  engineers  manage  all  quan>>es  (vs.  52  quan>ty  surveyors)  

Recommendation Don’t treat BIM as an isolated add-on à Create VDC methods for your work,

including § Performance targets § BIM § Revised workflow § Revised collaboration

BIM offers an integrated information basis

24

Image courtesy DPR

Integrated Concurrent Engineering is a method for collaboration

(c) 2014

1.  Develop Strategic Goals and Objectives for MEP Coordination

2. Organize a multi-disciplinary team for coordination

3. Co-develop performance and outcome objectives

4. Co-Develop Technical Logistics to manage coordination

5. Develop Pull Schedule to structure the work based on construction sequence

6. Manage against the performance objectives

Rethink your work processes given the combination of collaboration and BIM

(c) 2014

The IVL Method by Atul Khanzode, PhD Research advised by Martin Fischer, Glenn Ballard, and others

Set performance targets and track them Outcome  Metrics   Case  Study  1:   Case  Study  2:  Mechanical  Prefabrica>on  %   90%   30%  Plumbing  Prefabrica>on  %   90%   0%  Electrical  Prefabrica>on  %   40%   25%  RFIs  due  to  Conflicts  during  Construc>on   2  of  677   30  of  200  Number  of  Change  Orders  due  to  conflicts  during  Construc>on   0  of  311   30  of  230  Minutes  per  day  Superintendent  spent  resolving  issues  between  MEP  trades   20  -­‐  30   180  Average  Planned  Percent  Complete   80%   Did  not  track  %  Rework  Hours  compared  to  Total  Hours   Less  than  1%   20%  

(c) 2014

The Scientific Perspective “Science is knowledge which we understand so well that we can teach it to a computer; and if we don't fully understand something, it is an art to deal with it. Since the notion of an algorithm or a computer program provides us with an extremely useful test for the depth of our knowledge about any given subject, the process of going from an art to a science means that we learn how to automate something.” (Donald Knuth, Computer Programming as an Art, CACM, Dec. 1974)

“Automated  execu>on  of  processes  changes  everything.”  (Alan  Perlis,  1961)  

The  Business  Perspec>ve  

29  

•  Three  types  of  constraints  •  Precedence  •  Discrete  (Labor)  •  Disjunc>ve  (Workspace)  

•  Automated  scheduler    •  Varies  sequence  (thousands  of  viable  

schedules)  

•  Maximize  space  u>liza>on  

•  Eliminate  spa>al  clashes  

Tri-­‐Constraint  Method  (work  by  Rene  Morkos)  

30

Average bay occupancy 3.1%

3.3% 3.9%

4.6% 4.3%

3.0%

2.2%

3.3%

1.7%

0.0% 0.0% 0.5% 1.0% 1.5% 2.0% 2.5% 3.0% 3.5% 4.0% 4.5% 5.0%

Average Hourly Occupied Space (%)

Space is underutilized on some construction sites

Need a method to maximize work density

31  

Accessorize   BioEChemE  BioEChemE  

 SCM  schedule  duraQons  on  average  47%  shorter  than  LOB  

TCM  Basic  Results  

Men’s  Fashion  

TCM models labor resources  Accessorize  Project    

Schedule  dura>on  vs.  #  crews  

Men’s  Fashion  Project    Schedule  dura>on  vs.  #  crews  

BioEChemE  Project  Schedule  dura>on  vs.  #  crews  

Max  #  Crews  

Max  #  Crews  Max  #  Crews  

2014-15 CIFE Seed Research Project

Simulation-Based Approach to Accounting for Uncertainty and Variability in Look-Ahead Planning With Nelly Garcia-Lopez and James Choo

33

Motivating case: Curtain Wall Installation in 7-story office building in South San Francisco -  Critical path activity -  Opens up work for other trades (e.g., finishes) -  Disrupts ongoing work (6ft staging area around the perimeter) -  Vulnerable to variability -  Field managers were concerned about the installation crew

outpacing the fabrication crew

34

Source: Genzyme Corp http://www.sotawall.com/portfolio/United%20States/GenzymeCorporation-8568/

They tracked the fabrication and installation production and updated the chart daily

Fabrication Installation

35

0

100

200

300

400

500

600

700

800

-5 5 15 25 35 45 55 65 75 85

Num

ber o

f Uni

ts

Working days

Line of Balance View of Curtain Wall Fabrication vs Installation

5 Day Material Buffer

60

Planned production rate: 12 units/day

36

0

100

200

300

400

500

600

700

800

-5 5 15 25 35 45 55 65 75 85

Num

ber o

f Uni

ts

Working days

Subcontractor started fabrication earlier than planned

60 120

Planned production rate: 12 units/day

37

0

100

200

300

400

500

600

700

800

-5 5 15 25 35 45 55 65 75 85

Num

ber o

f Uni

ts

Working days

Actual fabrication rate was 22% slower than planned

60 120

39

Planned production rate: 12 units/day Actual production rate: 9.3 units/day

Installation crew will run out of material on day 39

Buffer was insufficiently sized to absorb the upstream variability

38

Identifying variability factors and tracking them is not sufficient to size buffers appropriately

39

Identify constraints & make ready

Determine sequencing and design operations

Break down processes into

activities for look-ahead window

Field managers correctly identified

constraints

•  PPC for curtain wall installation in previous project was 50%

•  Main reason for non-completion was predecessor activity (fabrication)

Look-ahead process (Hamzeh, Ballard & Tommelein 2011)

But were unable to design the operation to avoid variability impact

•  Variability was

underestimated •  Material buffer was

insufficient

Case summary: Construction managers want to manage variability but lack a formal method to do so

Aware of impact of variability

Intuitive management of

variability

No formal methods to analyze variability factors and predict

impact

Constraint checking during look-ahead planning

Create inventory buffer to shield installation from variability in fabrication

Will fabrication over/under-supply the site? How is installation affected?

40

Activity execution is affected by activity variability factors and schedule variability factors

Variability in execution: •  Start date •  Activity

duration (Ballard & Howell 1998, Thomas et

al. 2002, Tommelein et al. 1999)

41

Activity variability factors: •  Labor •  Tools & Equipment •  Materials and

components •  Information/plans •  Previous work •  Site conditions •  External

Activity execution affect

leads to

Schedule variability factors: •  Work in process

(Gonzalez et al. 2011) •  Site congestion

(Morkos et al. 2014)

Activity variability analysis needs to incorporate interdependencies between variability factors

Research method Quantitative data analysis (Kuhn & Johnson 2012)

1. Exploratory

data analysis 2. Confirmatory data analysis

3. Predictive modeling

•  Identify variables driving activity variability

•  Statistically confirm hypothesis resulting from part 1

•  Build model to predict variability impact on activity start and duration

42

Data acquired to date Activity tracking data collected daily at a hospital building project

by a CIFE partner over a period of 31 months (Nov 2011 – June 2014) •  30,000 total activity entries •  We cleansed the data-set:

•  25,170 activities entries with valid dates entered •  Manually classified into 761 activity types and Uniformat

categories

43

Data request We need activity tracking data for building projects that have implemented Last Planner: Data needed per activity: •  Activity Description •  Subcontractor/Team performing activity •  Planned start, planned finish, planned duration •  Actual start, actual finish, actual duration •  Reason for non-completion (category and root cause), reasons for

changes in start dates and duration •  Predecessors, successors (or schedule network) Please contact Professor Martin Fischer (fischer@stanford.edu) or Nelly Garcia-Lopez (ngarcial@stanford.edu) if you would like to be involved in this project.

44

Develop a unifying theory of project production management

Virtual vs. Physical Production Tradeoffs Automation Product-Organization-Process

Production Physics and Organizational Chemistry How to estimate capacity Multi-scale workflow examples Rapid learning cycles

Controllable Factors à Production Performance à Outcome Performance Optimization

What: EEE Performance How